From 820cd62fa17f3303ce8b518f3d9dd4dcefbd5e11 Mon Sep 17 00:00:00 2001
From: agra <alex@swipelab.co>
Date: Thu, 25 Jun 2026 13:57:56 +0300
Subject: [PATCH] docs: file issues 0185-0188

File four issue write-ups discovered alongside the 0179 work:
- 0185: binary-op operand auto-unwrap silently miscompiles a NULL ?T
- 0186: closure VALUE call does not coerce arg to ?T parameter
- 0187: lambda with inferred return type + block body with early returns
  mis-infers its return type
- 0188: closure-VALUE calls skip argument validation (arity + tuple spread)
---
 ...85-binop-optional-operand-silent-unwrap.md | 92 +++++++++++++++++++
 ...186-closure-optional-param-arg-coercion.md | 91 ++++++++++++++++++
 ...mbda-inferred-return-block-early-return.md | 55 +++++++++++
 ...-closure-value-call-arg-validation-gaps.md | 65 +++++++++++++
 4 files changed, 303 insertions(+)
 create mode 100644 issues/0185-binop-optional-operand-silent-unwrap.md
 create mode 100644 issues/0186-closure-optional-param-arg-coercion.md
 create mode 100644 issues/0187-lambda-inferred-return-block-early-return.md
 create mode 100644 issues/0188-closure-value-call-arg-validation-gaps.md

diff --git a/issues/0185-binop-optional-operand-silent-unwrap.md b/issues/0185-binop-optional-operand-silent-unwrap.md
new file mode 100644
index 00000000..03e6abba
--- /dev/null
+++ b/issues/0185-binop-optional-operand-silent-unwrap.md
@@ -0,0 +1,92 @@
+# 0185 — binary-op operand auto-unwrap silently miscompiles a NULL `?T` operand to garbage
+
+> **RESOLVED.** Root cause as diagnosed below: `lowerBinaryOp`
+> (`src/ir/lower/expr.zig`) auto-unwrapped optional operands UNCONDITIONALLY.
+> Fix: gate both operand unwraps on flow narrowing (the same `narrowed_refs`
+> mechanism issue 0179 introduced) — an un-narrowed `?T` operand is rejected via
+> the new `diagOptionalOperand`; a guard-narrowed operand still unwraps. `== null`
+> / `!= null` presence tests are unaffected (they return early before the
+> auto-unwrap). While fixing, an adversarial review of 0179 surfaced a real
+> soundness hole: `narrowed_refs` (keyed by per-function `Ref` index) leaked into
+> nested bodies whose `Ref` space overlaps (closure literals, generic/pack/comptime
+> monomorphization, AND — caught by an INDEPENDENT second-pass review — the JNI
+> native-method body path, where residue leaked between consecutive `#jni_main`
+> method stubs), letting an outer narrowed `Ref` falsely match a nested `Ref`.
+> Closed with `Lowering.NarrowGuard` (save/clear/restore around each such nested
+> body): lowerLambda (closure.zig), monomorphizeFunction (generic.zig),
+> createComptimeFunctionWithPrelude (comptime.zig), monomorphizePackFn (pack.zig),
+> synthesizeJniMainStub (ffi.zig). FnBodyReentry + the explicit clear in
+> lowerFunction (decl.zig) cover the rest. Regressions:
+> `examples/optionals/0921-optionals-binop-narrowing.sx` +
+> `examples/optionals/0922-optionals-binop-no-implicit-unwrap.sx`.
+>
+> **Discovered (separate, NOT fixed here):** calling a closure VALUE/variable
+> whose parameter is `?T` does not coerce the argument to the param type — a
+> concrete `7` arrives absent and `null` emits `ptr null` against a `{T,i1}`
+> param (LLVM verifier failure). Pre-existing, orthogonal to optional unwrap.
+> Filed as issue 0186.
+
+## Symptom
+
+An arithmetic / comparison binary op with an OPTIONAL operand (`a + b`,
+`a < b`, etc.) unconditionally unwraps the optional's payload — for a PRESENT
+optional it yields the value, for a NULL optional it yields the zero/garbage
+payload with NO diagnostic. Silent miscompile, same spirit as issue 0179 but a
+DIFFERENT code path: this auto-unwrap is in the binary-op lowering, NOT the
+`classify` / `coerceMode` coercion ladder that 0179 fixed.
+
+Per specs.md §Optional Types, the only legal ways to extract `T` from `?T` are
+`!` / `??` / `if v := opt` / pattern match / flow-sensitive narrowing after a
+`!= null` guard. There is no implicit unwrap-at-an-operand-position; a null
+operand must not silently become its zero payload.
+
+## Reproduction
+
+```sx
+#import "modules/std.sx";
+main :: () {
+  a : ?i64 = null;
+  b : i64 = 10;
+  c := a + b;            // prints "c = 10" (0 + 10) — silent miscompile, no diagnostic
+  print("c = {}\n", c);
+}
+```
+
+Expected: a compile error like 0179's
+(`cannot use a value of type '?i64' where 'i64' is expected: an optional does
+not implicitly unwrap; ...`), unless `a` is proven present by a `!= null`
+guard (flow narrowing), in which case the unwrap is sound.
+
+## Root cause
+
+`src/ir/lower/expr.zig` (`lowerBinaryOp`, ~line 3210, "Auto-unwrap optional
+operands for arithmetic/comparison"): both operand arms do an unconditional
+`.optional_unwrap` when the operand type is `.optional`, never reading the
+has_value flag and never consulting flow narrowing. This is the operand-side
+analogue of the coercion-side bug fixed in 0179.
+
+## Investigation prompt
+
+Gate the binary-op operand auto-unwrap on the SAME flow-narrowing mechanism
+0179 introduced (`Lowering.narrowed` / `narrowed_refs`, see
+`src/ir/lower/control_flow.zig` + `coerceMode`'s `.optional_unwrap` arm in
+`src/ir/lower/coerce.zig`):
+
+1. In `lowerBinaryOp`, when an operand is `?T`, only auto-unwrap it when its
+   lowered `Ref` is in `self.narrowed_refs` (proven present); otherwise emit
+   the same loud diagnostic 0179 uses and skip the silent unwrap. The operand
+   `Ref` is produced by `lowerExpr(bop.lhs/rhs)`, so a guard-narrowed local is
+   already tagged into `narrowed_refs` by `lowerIdentifier` — the gate is a
+   `narrowed_refs.contains(operand)` check.
+2. Decide the semantics for a present-but-mixed case (e.g. `?i64 + i64`): the
+   result stays `i64` (unwrap the optional operand) only when narrowed, as
+   above. Confirm comparisons (`==`/`!=`) against `null` are unaffected — those
+   are presence tests, not operand unwraps, and must keep working.
+3. Verify: the repro above must become a compile error; a guarded
+   `if a != null { c := a + b; }` must still compute `a + b` correctly; the
+   existing `examples/optionals/0900-optionals-optionals.sx` `guard2`
+   (`return a + b` after a compound `== null or` guard) must still pass.
+
+Add a positive regression (guarded arithmetic) + a negative regression
+(unguarded `?T` operand rejected), mirroring 0179's
+`examples/optionals/0919` / `0920`.
diff --git a/issues/0186-closure-optional-param-arg-coercion.md b/issues/0186-closure-optional-param-arg-coercion.md
new file mode 100644
index 00000000..6ac9bd17
--- /dev/null
+++ b/issues/0186-closure-optional-param-arg-coercion.md
@@ -0,0 +1,91 @@
+# 0186 — calling a closure VALUE with a `?T` parameter does not coerce the argument
+
+> **RESOLVED.** Root cause as diagnosed below: the closure-VALUE call path in
+> `src/ir/lower/call.zig` lowered args without coercing to the closure's
+> declared parameter types. Two-part fix: (1) `resolveCallParamTypes` now
+> returns the closure/function value's param types for an identifier callee
+> bound to a closure/fn VALUE in scope (so args lower with the right
+> `target_type`; precedes function-name resolution since a local value shadows
+> a function); (2) new free fn `coerceClosureCallArgs` coerces each
+> already-lowered user arg to the closure's param type via `coerceToType`,
+> applied at all three `call_closure` emission sites (local-variable callee,
+> struct-field callee, force-unwrap/expr callee) AND the local function-pointer
+> `call_indirect` path (which had the identical gap — an adversarial review
+> flagged that the `.function` branch of (1) typed fn-ptr args but never
+> coerced them). Now a concrete arg wraps present, `null` → absent — matching a
+> top-level fn call, for both closure values and fn-pointer values. Regression:
+> `examples/closures/0312-closure-optional-param-arg-coercion.sx` (local +
+> struct-field closure + fn-pointer value, concrete + null args).
+>
+> **Discovered while verifying (separate, NOT fixed here):** a lambda with an
+> INFERRED return type (no `-> T`) and a block body with early `return`s
+> mis-infers its return type (LLVM verifier failure) even with no optionals.
+> Filed as issue 0187. (The 0312 regression uses an explicit `-> i64` to avoid
+> it.)
+
+## Symptom
+
+When a closure value/variable (a `:=`-bound lambda, or any closure passed as a
+value) has a parameter of optional type `?T`, the call site does NOT coerce the
+argument to `?T`:
+
+- A concrete argument (`pick(7)`) is NOT wrapped to a present `?i64` — inside
+  the body the param reads as ABSENT (`p == null` is true), so the closure
+  silently returns the wrong branch.
+- A `null` argument (`pick(null)`) lowers `null` as a bare `ptr null` against a
+  `{i64, i1}` parameter, which fails LLVM verification:
+  `Call parameter type does not match function signature! ptr null  { i64, i1 }`.
+
+The SAME signature/body as a TOP-LEVEL function works correctly (e.g. issue
+0900's `guard`), so the bug is specific to the closure-VALUE call path
+(`src/ir/lower/call.zig`'s closure/fn-pointer call lowering), not optionals or
+flow narrowing. Found during the adversarial review of issues 0179 / 0185.
+
+## Reproduction
+
+```sx
+#import "modules/std.sx";
+norm :: (p: ?i64) -> i64 { if p == null { return -1; } return 99; }
+main :: () {
+  pick := (p: ?i64) -> i64 => {
+    if p == null { return -1; }
+    return 99;
+  };
+  print("pick 7: {}\n", pick(7));   // prints -1 (WRONG — should be 99; arg arrives absent)
+  print("norm 7: {}\n", norm(7));   // prints 99 (top-level fn, correct)
+  // print("pick null: {}\n", pick(null));  // LLVM verifier failure: ptr null vs {i64,i1}
+}
+```
+
+Expected: `pick(7)` prints `99` (the `7` wraps to a present `?i64`), and
+`pick(null)` compiles (the `null` lowers to an absent `?i64`), matching the
+top-level `norm`.
+
+## Root cause (hypothesis)
+
+The closure-value call path in `src/ir/lower/call.zig` lowers each argument and
+passes it to the closure's trampoline WITHOUT running the `coerceToType` step
+that the normal sx-to-sx call path applies against the callee's declared
+parameter types. So a `T → ?T` wrap (and `null → ?T`) never happens for a
+closure value's optional param. A top-level fn call coerces args to param types,
+which is why `norm` works.
+
+## Investigation prompt
+
+In `src/ir/lower/call.zig`, find the closure-value / fn-pointer call lowering
+(where `%cl.fn`/`%cl.env` trampolines are invoked — grep for `cl.fn` / the
+closure-call branch). Confirm it lowers args without coercing to the closure
+type's parameter types. The closure's parameter types are available from the
+`Closure(...)`/closure TypeInfo. Coerce each lowered argument to the
+corresponding parameter type via `self.coerceToType(arg, arg_ty, param_ty)`
+before building the call — mirroring the sx-to-sx call path. Verify:
+1. The repro above: `pick(7)` → `99`, `pick(null)` compiles and the body sees
+   absent.
+2. No regression in existing closure examples (`examples/closures/`).
+3. Add a regression `examples/closures/03xx-closure-optional-param.sx` covering
+   a concrete arg (wraps present) and a `null` arg (absent) into a closure-value
+   `?T` param.
+
+Note: this is purely an argument-coercion gap at the closure-value call site;
+it is unrelated to the implicit-optional-unwrap family (issues 0179 / 0185),
+which is already fixed.
diff --git a/issues/0187-lambda-inferred-return-block-early-return.md b/issues/0187-lambda-inferred-return-block-early-return.md
new file mode 100644
index 00000000..c9ff2e55
--- /dev/null
+++ b/issues/0187-lambda-inferred-return-block-early-return.md
@@ -0,0 +1,55 @@
+# 0187 — lambda with INFERRED return type + block body with early `return`s mis-infers its return type (LLVM verifier failure)
+
+## Symptom
+
+A `:=`-bound lambda (closure literal) that has NO explicit `-> T` return type
+and whose body is a BLOCK containing `return` statements infers the WRONG
+return type (apparently `void`). Calling it and using the value fails LLVM
+verification (`Call parameter type does not match function signature! ... i64
+undef` / `Function arguments must have first-class types!`). Adding an explicit
+`-> T` makes it work. No optionals or flow narrowing are involved — found while
+verifying issue 0186.
+
+## Reproduction
+
+```sx
+#import "modules/std.sx";
+main :: () {
+  // inferred return type, block body with early returns — NO optionals
+  f := () => { if 1 > 0 { return 11; } return 22; };
+  print("f: {}\n", f());   // LLVM verification failed (return type inferred void/undef)
+}
+```
+
+Workaround / contrast (works): annotate the return type —
+`f := () -> i64 => { if 1 > 0 { return 11; } return 22; };`
+
+## Root cause (hypothesis)
+
+The lambda return-type inference in `lowerLambda` (`src/ir/lower/closure.zig`,
+the `ret_ty` computation around line 164: `const inferred =
+self.inferExprType(lam.body);`) does not infer the type from the body's
+`return` statements when the body is a block. For a block whose value is
+produced only via early `return`s (not a trailing tail expression),
+`inferExprType` likely yields `.void`, so the lambda is built with a void
+return while the body actually returns `i64` — the mismatch surfaces at the
+call site / LLVM verifier.
+
+## Investigation prompt
+
+In `src/ir/lower/closure.zig`, the lambda return-type inference path
+(`inferExprType(lam.body)` ~line 164) must, for a block body, infer the return
+type from the body's `return` statement operands (matching how
+`lowerValueBody` / the function-decl return inference handles bodies with early
+returns), not just the block's tail-expression value. Reuse the existing
+return-type inference the top-level fn path uses (a top-level
+`f :: () { if c { return 11; } return 22; }` with inferred return works — see
+why, and apply the same to lambdas). Verify:
+1. The repro prints `f: 11`.
+2. `examples/optionals/0919`/`0921` and `examples/closures/0312` still pass
+   (0312 deliberately uses explicit `-> i64` to dodge this bug — once fixed, an
+   inferred-return variant should also work).
+3. Add a regression `examples/closures/03xx-lambda-inferred-return-early.sx`.
+
+Unrelated to the optional-unwrap family (0179/0185) and the closure-arg
+coercion fix (0186); purely lambda return-type inference.
diff --git a/issues/0188-closure-value-call-arg-validation-gaps.md b/issues/0188-closure-value-call-arg-validation-gaps.md
new file mode 100644
index 00000000..728111d1
--- /dev/null
+++ b/issues/0188-closure-value-call-arg-validation-gaps.md
@@ -0,0 +1,65 @@
+# 0188 — closure-VALUE calls skip argument validation: no arity check + runtime-tuple spread not expanded
+
+## Symptom
+
+Calling a closure VALUE (a `:=`-bound lambda, struct-field closure, fn-pointer
+value) does NOT validate arguments the way a top-level function call does. Two
+distinct gaps, both pre-existing (surfaced during the adversarial review of the
+issue-0186 fix; 0186 fixed only arg COERCION for correctly-counted args):
+
+1. **No arity check.** A closure value called with the WRONG number of args
+   compiles and silently drops/ignores extras (or reads garbage for missing
+   ones), exit 0. A top-level fn call diagnoses arity.
+2. **Runtime-tuple spread `f(..tuple)` is never expanded for a closure value.**
+   The spread leaves a `Ref.none` placeholder (`call.zig` ~line 404) that the
+   `call_closure` sites emit as `undef`, so the call passes garbage.
+
+## Reproduction
+
+```sx
+#import "modules/std.sx";
+main :: () {
+  // (1) arity: extra arg silently dropped
+  one := (a: i64) -> i64 => a;
+  print("arity: {}\n", one(1, 2));      // prints 1 (no error; `2` dropped)
+
+  // (2) spread into a closure value → garbage
+  add := (a: i64, b: i64) -> i64 => a + b;
+  pair := (10, 20);
+  print("spread: {}\n", add(..pair));   // prints garbage (e.g. -17590042754976), not 30
+}
+```
+
+Expected: (1) an arity diagnostic (as for a top-level fn); (2) `add(..pair)`
+expands to `add(10, 20)` → `30`, OR a clear diagnostic that spread into a
+closure value is unsupported (never silent garbage).
+
+## Root cause (hypothesis)
+
+The closure-value call paths in `src/ir/lower/call.zig` (the three
+`call_closure` emission sites + the local `call_indirect` fn-pointer path) build
+the arg list and emit directly without (a) an arity check against
+`closure.params.len` / `function.params.len`, and (b) without running the
+runtime-slice/tuple spread expansion that the normal call path uses
+(`packVariadicCallArgs` / the `Ref.none` spread placeholder is never resolved
+for closures). The pack-spread `..xs` path (`packSpreadRefs`) handles comptime
+packs but not a runtime tuple value spread into a closure.
+
+## Investigation prompt
+
+In `src/ir/lower/call.zig`, for each closure-value / fn-pointer-value call site
+(grep `call_closure` and the local `call_indirect` path ~line 655):
+1. Add an arity check against the callee value's `closure.params` /
+   `function.params` length (mirror `checkCallArity` used for top-level fns),
+   accounting for the implicit `__sx_ctx` slot.
+2. Either expand a runtime-tuple/slice spread argument into positional args for
+   closure values (as the normal call path does), or emit a located diagnostic
+   that spread into a callable value is unsupported — never emit the `Ref.none`
+   placeholder as `undef`.
+3. Regression: extend `examples/closures/0312-...` or add
+   `examples/closures/03xx-closure-value-arity.sx` covering both the arity
+   diagnostic and the spread behavior.
+
+Unrelated to the arg-COERCION fix (issue 0186, already landed) — that fix
+correctly coerces a correctly-COUNTED arg; these gaps are about COUNT and
+spread expansion.